Embed Size (px)
Citation preview
R-matrix Analysis (II)
GANIL TALENT SchoolTALENT Course 6
Theory for exploring nuclear reaction experiments
GANIL 1st-19th July
Ed Simpson
University of Surrey
R-matrix Analysis (II)
• Tutorial so far:
– Setup the particle pairs
– Added states to compound nucleus
– Identified spins and parities of states
– Fit the widths and energies to experimental data
• Background resonances
• Fitting with MINUIT - examples
• Sensitivity to R-matrix radius parameter
• Error analysis with MINUIT
Background resonances
• Infinite expansion, infinite number of levels
• In principle, one pole for each Jπ
• Simulate non-resonant contributions…
– Higher lying resonances excluded in the fit
– (Weak) Non-compound nucleus (direct) mechanisms
– Corrections to hard-sphere phase shift in scattering
• For reaction channels, usually only a few Jπ are important
• For scattering, all Jπ may be required
12C(α, α)12C
12C(α, α)12C
Probable background from a specific resonance
Ex=8.67 MeV Ex=13.17 MeV
Sp(16O) Sα1(16O)
....................................................................................................................................................
Sensitivity to R-matrix radius
Sensitivity to R-matrix radius
Sensitivity to R-matrix radius
Radius E (MeV) Γ (keV) E (MeV) Γ (keV)
5.0 9.5803 415 10.3563 31.7
5.5 9.5764 392 10.3569 30.4
6.0 9.5659 376 10.3567 29.8
NNDC 9.585 420±20 10.356 26±3
a = 1.5 (21/3 + 121/3) = 5.32
Fitting in R-matrix
• Chi-squared minimization using MINUIT2 MINUIT2 Fitting • You are far smarter than the fitting routine! • MINUIT is very effective at optimizing the parameters if the
starting set is good... • ... But it doesn’t understand the spectrum and the physical
significance of parameters • The essential elements must be all in place - energies,
widths and interferences must be approximately correct • Start simple and build up the calculation
MINUIT fitting 1
MINUIT fitting 2
Parameter errors
• After the fit we have an optimal set of P parameters {p0}
• Taylor expand the chi-squared about the fitted parameters:
See e.g. F. James, Statistical Methods in Experimental Physics
and Filomena’s Lectures from last week
Parameter errors and correlations
• One sigma variation in the probability is given by:
• This defines multi-dimensional ellipses in parameters space around the minimum
• Parameter correlations defined as:
See e.g. F. James, Statistical Methods in Experimental Physics
and Filomena’s Lectures from last week
Parameter correlation matrix
Parameter List
j=1_la=1_energy ( 0) 9.585402e+00 Fitted by Minuit
j=1_la=1_ch=1_rwa ( 1) 7.826643e-01 Fitted by Minuit
j=2_la=1_energy ( 2) 1.500000e+01 Fixed
j=2_la=1_ch=1_rwa ( 3) 3.922930e+03 Fitted by Minuit
j=3_la=1_energy ( 4) 1.035718e+01 Fitted by Minuit
j=3_la=1_ch=1_rwa ( 5) 4.609894e-01 Fitted by Minuit
Correlation Matrix
0 1 3 4 5
0 1.000000 0.496720 -0.000224 0.087533 -0.135543
1 0.496720 1.000000 -0.000294 0.136914 -0.217443
3 -0.000224 -0.000294 1.000000 -0.000075 -0.000354
4 0.087533 0.136914 -0.000075 1.000000 -0.092854
5 -0.135543 -0.217443 -0.000354 -0.092854 1.000000
9.585 1-
15.000 3-
10.353 4+
MINOS Error analysis: output/covariance_matrix.out
Other capabilities...
• Capture reactions – External part of capture amplitude can be included – Explicitly add the final state as a level, set the ANC as you
would a width
• Experimental effects – Beam convolution – Gaussian distribution of incident
energies – Target integration – thickness of target – Added on per segment basis (different experiments,
different targets)
• Reaction rate calculations – Based on an R-matrix calculation, calculate an
astrophysical reaction rate – Computationally intensive (slow...)
Summary
Phenomenological R-matrix fitting provides one method to understand the structure of nuclei at low excitations. It provides a means to calculate low-energy scattering and reaction cross sections for nuclear astrophysics, materials analysis and other applications. R-matrix calculations should begin as simply as possible, identifying the strongest states, before adding weaker contributions. Fitting using MINUIT should only be done once the key elements are established. In principle the R-matrix fits should be weakly dependent on the R-matrix radius parameter, but this should be checked.
